Ontology-based knowledge graph infrastructure for interoperable atomistic simulation data

A team of researchers has proposed a new ontology-based infrastructure designed to represent and integrate atomistic simulation data as a knowledge graph, as detailed in a recent preprint paper published on the arXiv server. This work, announced on April 26, 2024, addresses the critical challenge of data reuse in computational materials science and chemistry, which is often hampered by incompatible file formats, missing metadata, and non-standardized records of scientific workflows and data provenance. The core innovation of the proposed system lies in its combination of specialized domain ontologies—formal frameworks that define concepts and relationships within a field—with a practical software framework. This dual approach allows the infrastructure to capture data from two primary sources: vast existing repositories of legacy simulation results and, crucially, directly from simulation software as new calculations are performed. By structuring this diverse information into a unified knowledge graph, the system creates a machine-readable and semantically rich network where data points are interconnected based on their scientific meaning, rather than just their file format. This development represents a significant step toward achieving the long-sought goal of true interoperability in computational science. For researchers, it promises to dramatically accelerate discovery by making previously siloed or unusable data easily searchable, comparable, and reusable for new studies, such as training machine learning models or validating novel simulation methods. The infrastructure effectively acts as a universal translator and filing system for the complex, multi-step processes involved in simulating materials and molecules at the atomic level, ensuring that the full context of how a data point was generated is preserved and accessible.